Is Census Data Shorting Total Retirement Wealth?

Commentators are questioning the accuracy of Census Bureau data that examines how well off Americans are in retirement.

 A recent issue brief from the Center for Retirement Research at Boston College (CRR) asks if retirement income flow data captured by the Census Bureau’s Current Population Survey (CPS) understates retirees’ actual wealth. The analysis compares income figures in the CPS with those reported to the Internal Revenue Service and the United States Federal Reserve, looking closely at households with retired heads age 65 to 84.

According to CRR researchers, all the comparisons confirm that the CPS dramatically underreports total retirement wealth, especially when it comes to expected income from 401(k) and individual retirement accounts (IRAs) for those at higher income levels.

Importantly, the CRR’s analysis finds the impact of this underreporting is most substantial for Americans in the upper two income quintiles, although it is experienced to some degree across the entire income spectrum. The CPS findings are most accurate for middle-income families, the analysis contends, who according to the CRR hold relatively little wealth from workplace retirement plans and IRAs.

The Federal Reserve data comes from the 2013 Survey of Consumer Finances (SCF), a nationally representative survey of about 6,100 households, while the IRS data is derived from tax information. The data discrepancy between these sources is not subtle, according to the CRR. For example, the CPS reports $18 billion of total 401(k)/IRA income in 2012, while the SCF reports $220 billion and the IRS $229 billion.

The differences in reported income arise for several reasons, the CRR analysis finds. First, individuals may not withdraw money that is available to them, meaning income in a given year may present an artificially low picture of total wealth. Many 401(k) and IRA holders do not withdraw money until their early 70s, researchers observe, when they become subject to the IRS’s required minimum distribution rules. 

Second, according to the CRR analysis, questions about retirement wealth included in the CPS questions put a potentially misleading emphasis on “regular” payments from retirement accounts, leading to confusion about whether to count things like as-needed or lump sum withdrawals as income.

As defined contribution plans become more prevalent as a source of income for retirees, the CPS’s failure to capture income from 401(k)s and IRAs will lead to increasing underreporting of retirement resources, the CRR says. As the discrepancy grows, the CRR warns that researchers and Americans generally will lose confidence in the CPS. This loss of confidence is serious, the analysis concludes, given the value of the survey in terms of providing significant annual data with a large sample size broken down according to a number of demographic variables.

CRR researchers say the federal government has responded with plans to test different question patterns to try to improve the accuracy of data on retirement income. For its part, the CRR analysis urges the CPS to include a measure of total 401(k)/IRA assets (i.e., not just assets claimed as income in a given year) both to verify income estimates and to calculate “potential” defined contribution income in cases when households are not drawing on their retirement wealth.

The full CRR analysis is available here.